README.md

---
title: MuseControlLite
emoji: 👁
colorFrom: red
colorTo: blue
sdk: gradio
sdk_version: 6.1.0
app_file: app.py
pinned: false
license: mit
short_description: Inference for Stable-Audio-Open with more controls
---

## MuseControlLite (Space)

Gradio UI for MuseControlLite adapters on top of `stabilityai/stable-audio-open-1.0`.

### Requirements
- **GPU Space** is required for generation (fp16 by default).
- A Hugging Face token with access to `stabilityai/stable-audio-open-1.0` (set as a Space secret, e.g., `HF_TOKEN`).

### What happens on startup
1) Installs Python deps from `requirements.txt` (includes `gradio`, `gdown`, `diffusers` fork).  
2) Downloads MuseControlLite checkpoints with  
   `gdown 1Q9B333jcq1czA11JKTbM-DHANJ8YqGbP --folder`  
   into `checkpoints/` if they are missing.

### Using the Space
1) Provide a text prompt.  
2) Upload a 47.5s (or longer) audio file when using MuseControlLite conditions.  
3) Select condition types (`melody_stereo`, `melody_mono`, `dynamics`, `rhythm`, `audio`) and adjust guidance/scales if needed.  
4) Click **Generate**. Output is a single 47.5s WAV plus a short status summary.

### Tips
- `melody_stereo` cannot be combined with `dynamics`, `rhythm`, or `melody_mono`.  
- For audio in/out-painting, use the audio condition with the masking sliders.  
- Default examples are preloaded in the UI for quick tests.

### Local run (optional)
```bash
pip install -r requirements.txt
gdown 1Q9B333jcq1czA11JKTbM-DHANJ8YqGbP --folder
huggingface-cli login
python app.py
```

### Acknowledgments
- Original repository: https://github.com/fundwotsai2001/MuseControlLite